A method for GPS navigation which uses an interacting multiple-model (IMM) estimator with a probabilistic data association filter (PDAF) improves navigation performance. The method includes (a) providing two or more models of GPS navigation, with each model characterized by a model state vector which is updated periodically, (b) providing for each model a corresponding filter for deriving, for each period, a current value for the corresponding model state vector based on current measurements made on parameters affecting the corresponding state vector; and (c) applying an interacting multiple model (IMM) estimator to provide, for each period, a current value for a system state vector using the current values of the model state vectors for that period and their corresponding filters. Each model state vector may include one or more of the following: variables: 3-dimensional position, 3-dimensional velocity, satellite clock bias, satellite clock drifts and one or more other satellite parameters. The current value of the system state vector may be a weighted average of the current values of the model state vectors, where the weights are a set of mode probabilities. In addition, one or more of the filters is a probabilitic data association filter (PDAF).
Legal claims defining the scope of protection, as filed with the USPTO.
1. A method for performing GPS navigation in a GPS receiver, comprising: providing a plurality of models of GPS navigation, each model characterized by a model state vector which is updated periodically; providing for each model a corresponding probabilistic data association filter (PDAF) for deriving, for each periodic update, a current value for the corresponding model state vector based on a current observation made on parameters affecting the corresponding state vector, wherein each PDAF determines whether the current observation is a correct observation or a false observation such that each PDAF updates the corresponding model state vector if the current observation is a correct observation and does not update the corresponding model state vector if the current observation is a false observation; and applying an interacting multiple model (IMM) estimator to provide, for each period, a current value for a system state vector using the current values of the model state vectors for that period and their corresponding PDAFs.
2. A method as in claim 1 , wherein each model state vectors includes one or more of the following variables: 3-dimensional position, 3-dimensional velocity, clock bias, and clock drift.
3. A method as in claim 1 , where the current value of the system state vector is a weighted average of the current values of the model state vectors and wherein a set of mode probabilities are assigned as weights for the weighted average.
4. A method as in claim 3 , wherein the IMM estimator calculates mixing probabilities for each model based on previous measurements made on the parameters.
5. A method as in claim 4 , wherein for each cycle, prior to deriving the current values of the model state vectors in the filters, using the mixing probabilities, the IMM estimator calculates an initial valve for each model state vector by weighting the current model state vectors of the immediately previous cycle.
6. A method as in claim 4 , using the current values of the model state vectors derived by the filters, the IMM estimator calculates the mixing probabilities for use in the next cycle, and the mode probabilities for providing the current value of the system state vector.
7. A method as in claim 1 , wherein each PDAF applies a statistical difference between the correct observation and the false observation to derive the current value of the model state vector.
8. A method as in claim 7 , wherein the statistical difference is expressed by a first probability density function for the correct observation, and a second probability density function for the false observation.
9. A method as in claim 8 , wherein the first probability density function and the second probability density function are each a function of a distance in standard deviations between the current observation and its nominal value.
10. A method as in claim 8 , wherein the first probability density function and the second probability density function are each a function of a measured value for the parameters affecting the model state vector.
11. A method as in claim 8 , wherein the first probability density function comprises one of: a Gaussian distribution, a Rayleigh distribution or an exponential distribution.
12. A method as in claim 8 , wherein the second probability density function comprises one of: a uniform distribution, a linear distribution, an exponential distribution, or a combination of a uniform distribution and a linear distribution.
13. A method as in claim 1 , wherein one of the models pertains to multipath effects.
14. A method as in claim 1 , wherein one of the models pertains to a signal-to-noise ratio.
15. A method as in claim 1 , wherein one of the models pertains to a GPS receiver moving at constant velocity.
16. A method as in claim 1 , wherein one of the models pertains to motions of non-constant velocities.
Cooperative Patent Classification codes for this invention. Click any code to explore related patents in that topic.
April 5, 2007
June 23, 2009
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.